The use of unmanned ground vehicles (UGV) has become wide-spread due primarily to the ability of such vehicles to travel into unknown and hazardous environments that otherwise can only be reached by manned vehicles with high safety risks and/or operational costs. Depending on operational requirements and cost constraints, unmanned ground vehicles can be designed with different levels of autonomy. The level of autonomy is usually measured by the frequency of human monitoring and assistance. Unmanned ground vehicles with high levels of autonomy are generally equipped with a guidance system that can autonomously generate appropriate vehicle guidance commands to lead the unmanned vehicles safely to their destinations while avoiding obstacles detected during traversal. As a result, these unmanned vehicles can continue to operate autonomously for extended periods of time.
Unmanned Ground Vehicles operating in outdoor unstructured or off-road environments must often deal with previously unknown obstacles discovered by their onboard perception sensors during traversal. Consequently, it is desirable that the motion planner residing in the guidance system be able to replan a safe path to lead the unmanned ground vehicle away from newly detected obstacles. In addition, future unmanned ground vehicles may be required to traverse at speeds significantly higher than possible in current unmanned ground vehicles, especially in applications where such vehicles must operate alongside manned vehicles or must reach remote destinations. However, when traveling at high speeds, the guidance system of an unmanned vehicle has only a very limited time to perform replanning. Therefore, rapid motion replanning is essential to avoid moving in the undesirable direction and risking collision with obstacles.
In general, motion planning for long-range missions can be divided into two stages: pre-mission path planning and real-time motion planning. During the pre-mission path planning stage, a reference path (position only) is determined that is free of obstacles based on a priori information such as maps, satellite images, and the like. After the mission has begun, real-time motion planning utilizes the reference path determined during pre-mission path planning to generate high-level guidance commands such as position, velocity, heading, and the like, that can be safely followed by the vehicle control system.
Conventionally, local path planning is used to revise the reference path determined during the pre-mission path planning to avoid obstacles detected en route. This is primarily because local path planning is simple to implement and works well in simple obstacle settings, such as where the ground vehicle must maneuver around a small obstacle. However, local path planners are vehicle-centric and only use information provided by the vehicle's onboard perception sensors about the vicinity of the ground vehicle. In addition, local path planners do not carry terrain information and do not register obstacle geo-locations in a global coordinate system. Thus, local path planners are incapable of negotiating complex obstacle settings such as a blocked urban canyon, a destroyed bridge, or the like. In such instances, the local path planner will be unable to revise the reference path to generate an obstacle-free path. Consequently, an unmanned ground vehicle employing only local path planning in combination with pre-mission path planning will be trapped because the vehicle's guidance system cannot find a way past the obstacle to the vehicle's destination.
Consequently, it would be desirable to provide a motion planner for an unmanned ground vehicle that is capable of traversing at high speeds in partially known environments while negotiating complex obstacle settings. Such a motion planner should construct high-level guidance commands such as position, velocity, heading, and the like, that are within the capability of the vehicle control system (i.e., guidance commands can be followed by the vehicle control system), lead the unmanned ground vehicle to its destination while avoiding obstacles, and minimize the cost of traversal.